15:00:39 #startmeeting gantt 15:00:41 Meeting started Tue Sep 16 15:00:39 2014 UTC and is due to finish in 60 minutes. The chair is bauzas. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:42 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:44 The meeting name has been set to 'gantt' 15:00:47 o/ 15:00:53 n0ano: oops 15:00:58 #chair n0ano 15:00:59 Current chairs: bauzas n0ano 15:01:05 \o 15:01:08 NP, you can run it\ 15:01:24 n0ano: well, I'm not sure we will have people here 15:01:32 hi 15:01:38 awesome 15:01:57 I just didn't think there was much to discuss this week and I don't like meetings for meetings sake but if there's people who want to talk that's fine 15:02:16 n0ano: yeah I know 15:02:22 n0ano, did jay do his write-up? 15:02:34 jaypipes is not here yet 15:02:41 haven't seen it yet, I pinged him yesterday and haven't heard back yet 15:02:49 PaulMurray: and I don't think he did had time 15:03:24 bauzas, I know the feeling - its been a hard week for me - will lighten up and looking forward to contributing a bit more 15:03:34 PaulMurray: no worries 15:03:52 PaulMurray, theres ~1000 bugs we can all look at :-) 15:03:56 PaulMurray: as Nova is in FF, all the patches are just for working on Kilo 15:04:27 PaulMurray: that said, as the level of changes is really high, these extra weeks are worth it 15:04:34 ok, let's start 15:04:39 #topic Gantt status 15:04:54 so, we all were on previous meeting 15:05:26 if someone lurking here and wants to know what's the plan, let's ping here 15:06:07 ok, assuming everybody is aware that we decided to work on objectify the dicts 15:06:21 and provide claims on Scheduler too 15:06:59 so, most of my week is currently busy on https://review.openstack.org/#/c/119807/ 15:07:12 bauzas, is any objects work going to get through in FF? 15:07:33 PaulMurray: all the patches will have to be merged by Kilo 15:07:45 PaulMurray: unless it's a critical bug 15:07:59 PaulMurray: I'm not sure we can land any patch by Juno now 15:08:13 +1 15:08:14 PaulMurray: I know you worked on the ComputeNode object 15:08:19 bauzas, that's what I thought - just checking 15:08:47 all part of developement will be faster once we split out gantt 15:08:51 PaulMurray: yep, understood. I'm just saying that ComputeNode object is far from being ready for being used 15:09:19 by saying that, I mean that this object is missing some fields and has a nested object too 15:09:44 bauzas, you had started a patch on CN - do you want me to take it over 15:09:51 so, as I said, I'm working on removing the nested Service object, and will have to work on adding the missing fields, like the pci_devices fields IIRC 15:10:14 PaulMurray: any help is welcomed, in particular as you worked on the CN object yet 15:10:32 PaulMurray: the above patch ^ is about removing the Service object dependency 15:11:06 PaulMurray: most of the work is already done except unittests modifications, I wanted to see if Tempest was happy or not 15:11:46 PaulMurray: your help will probably be good if you could resume my previous work on the missing fields 15:12:09 bauzas, ok , will do 15:12:09 https://review.openstack.org/97837 - Add missing fields to ComputeNode object 15:12:37 ^here I tried to look at what was missing in the CN object 15:12:44 of course, it needs a huge rebase 15:13:19 as you understand both, these tasks are preliminary but mandatory for using the CN object 15:13:53 bauzas, I'll do it - I'll split it to a patch for each field 15:14:03 once the CN object being fixed (hoping in early Kilo), we will be able to modify the update_resource_stats() method and pass an Object instead of a stats dict 15:14:11 PaulMurray: +1 15:14:12 bauzas, the supported instances is done wrong in ironic and needs special attention 15:14:23 PaulMurray: could you just rebase on my above patch too ? 15:14:35 PaulMurray: because I'm changing the CN object alot 15:14:49 PaulMurray: that Service dependency was pretty strong 15:15:10 PaulMurray: and it was impossible to cut it without modifying all implicated methods 15:15:23 bauzas, ok - meaybe we can talk tomorrow morning about how to coordinate that 15:15:29 strong +1 here 15:15:31 bauzas, don't want to send n0ano to sleep 15:15:39 :) 15:15:45 huh, what, did you say something :-) 15:15:58 eh, I hope that n0ano is enough caffeinated 15:16:09 working on my second cup, I'm good 15:16:15 awesome 15:16:29 so, that's basically the plan 15:16:44 I'm really worried about the required level of changes needed for having the CN object usable 15:17:01 that means we absolutely need to get these changes by Kilo-1 15:17:19 if we want to work on the next steps 15:17:48 wrt all the talks about the Kilo process, I'm really concerned about the timeframe and the support we'll get for Kilo 15:17:55 in particular if slots are implemented 15:18:29 I'm not a believer in slots, I think that's just a coarse way to slow down progress but we'll see 15:18:48 n0ano: maybe you know my opinion... 15:18:59 n0ano: yeah, I'm -2 on slots 15:19:19 but that's not the place here for discussing that :) 15:19:50 good news is I have people who want to work on gantt so we have some development resources available, we just need well defined, specific tasks for them. 15:19:51 #action PaulMurray and bauzas to coordinate on ComputeNode modifications needed for the split 15:20:24 n0ano: based on my previous experience with Juno, we will perhaps need more involvement than just committers 15:20:39 n0ano: but any help is good to take 15:20:54 n0ano: long story short, we don't need committers, we need reviewers 15:21:15 I think we have a silent majority of people who want gantt so we have more support than you think, core reviewers unfortunately will be an issue 15:21:29 n0ano: that's my thought 15:21:39 n0ano: we don't have a contributors issue 15:22:28 n0ano: and if we have slots in Kilo, we will only have core support for a certain amount of time 15:22:35 anyway 15:22:47 any other things to discuss about the split ? 15:23:06 can I move forward and raise the open topic ? 15:23:32 silent majority, any question to ask ? 15:23:34 go for it 15:23:53 #topic Open discussion 15:23:59 so 15:24:17 just wanted to make sure everybody is aware of what will happen for the Summit ? 15:24:28 ie. the new Summit agenda ? 15:24:44 by Summit, I mean "Design Summit" of course 15:24:48 what will happen? 15:24:51 I thought we were still debating that 15:25:12 n0ano: I think the main consensus has been reached, hence my discussion 15:25:24 then what's your take on it 15:26:26 #link http://lists.openstack.org/pipermail/openstack-dev/2014-September/045844.html 15:26:29 PaulMurray: ^ 15:26:51 thanks 15:26:52 n0ano: I want to make sure everybody can review the proposals for Day 2 and 3 15:27:02 so basically let me summarize 15:27:14 Day 1 will be for cross-projects discussion 15:27:21 => Gantt can claim for a call here 15:27:35 +1 15:27:41 indeed 15:27:53 Day 2 and Day 3 will be about project important discussions 15:28:18 https://etherpad.openstack.org/p/kilo-nova-summit-topics is the etherpad for tracking these topics 15:28:23 for Nova I mean 15:28:47 so, while we could discuss on Gantt for the day 1, we will only discuss about the sched split in Day 2 and Day 3 15:28:59 ie. what should be the scheduler in the next future 15:29:23 PaulMurray: feel free to review the above etherpad and leave comments if so 15:29:30 ok 15:29:38 a little cart before the horse, we really need to work out the split before we get concerned about future uses 15:29:54 PaulMurray: I just put the motto "scheduler split" and I'm happy to see debates about it there 15:30:04 n0ano: agreed 15:30:25 n0ano: that said, I think we need to at least restate our concern 15:30:37 +1 15:30:56 n0ano: I'm still wondering how Cinder or Neutron are scheduling 15:31:01 this also keeps the scheduler issues at the front of everyone's mind 15:31:15 n0ano: any guidance from these folks would be good here 15:31:29 cinder basically re-implemented the nova scheduler, I talked to the Cinder PTL, he is completely on board with gantt 15:31:49 n0ano: here comes Neutron... 15:32:04 (the cinder PTL is lives 30 miles away from me, we had coffee together one morngin) 15:32:14 n0ano: that said, a very quick glance on the Cinder code didn't showed me same classes or the same RT model 15:32:42 * bauzas wishes having Teleportals like in Star-Trek 15:32:46 they did their re-implementation a long time ago, the two schedulers have divered 15:32:55 n0ano: that's my understanding 15:32:56 s/divered/diverged 15:33:10 n0ano: so that's not an easy backport 15:33:24 but their goals are close enough that a single scheduler should work 15:33:26 at least, we need to know their concepts 15:33:40 anyway 15:33:48 back to the Summit discussion 15:33:55 Day 4 is still obscure to me 15:33:56 I believe nova sched is a superset of cinder, converging should be do able, but anyway 15:34:26 this sounds like the theme of day 1 15:34:44 understanding projects' needs 15:35:01 Day 4 seems to be a non-formal mode of discussions 15:35:22 ie. no time, just a big meetup 15:35:24 day 4 sounds as if it's like minded people get together 15:35:33 yeah, like BoFs or pods 15:35:56 e.g. just the gantt people on day 4, everyone argues things on day 2&3 15:36:14 I was seeing this as a possible use of pods 15:36:19 day 4 is just a big pod 15:36:37 looks like day 1-3 = discussion, day 4 = decide actions 15:36:41 seems like we'll be stuffing a `lot` of discussions in days 2&3 15:36:42 ie. people can come in and leave 15:37:00 n0ano: I agree 15:37:06 PaulMurray: I don't think so 15:37:13 bauzas, no? 15:37:39 PaulMurray: I'm seeing day 2 and day 3 as the same kind of talks we had from the previous summits (ie. a 60-min topic debated) 15:37:49 PaulMurray: where actions can be taken 15:38:01 PaulMurray: while day 4 is just a pure brainmind 15:38:20 PaulMurray: ie. something for discussing what's coming in on a non-formal mode 15:38:53 bauzas, I guess it depends on the topic 15:38:55 so, about what can be related to Gantt, I think it's covered by the etherpad for day 2-3 15:39:10 that said, I can't see any info about day 1 15:39:18 bauzas, I was thinking about the api versioning topic last time 15:39:23 so, if we want to talk about Gantt in day 1, we need to make sure we'll be able to 15:41:11 ok, I think I'm done on my side 15:41:22 any other things to discuss ? 15:41:44 I'm good 15:41:59 k 15:42:06 closing the call then 15:42:08 see you all 15:42:14 bye 15:42:14 tnx everyone 15:42:19 PaulMurray: feel free to ping me tomorrow 15:42:23 #endmeeting